JPS61153785A - Vertex detecting method - Google Patents

Abstract

PURPOSE:To detect a correct vertex by obtaining an angle difference between mean vectors before and after a noticing point on a curved line and detecting this noticing point as the vertex of the curved line when the angle difference exceeds a predetermined value. CONSTITUTION:A vector adder 110 change a direction code held by stages 105, 106, 107 into a vector expression and adds to obtain a mean vector Sc. A differential device 113 calculates an angle difference between the mean vector Sc and a vector Cn of a noticing point. In a comparator 115, the angle difference thetais compared with a predetermined value Th1 and if theta>Th1, a switch 116 is operated, and an output (Sc) of a vector adder 110 and the output of the other vector adder 111 are outputted to a difference device 113. In the difference device 113, the angle difference ¦ScEc¦ is obtained and compared with a predetermined value Th7 by the compara tor 115. When the angle difference is larger than Th7, the noticing point is detected as a vertex.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a vertex detection method suitable for linear approximation of a curve.

[Prior art]

Figure 4(a) is a curve expressed as a chain code,
Each number is a direction code and is a vector shown in FIG. Conventionally, a first-order method is used to find the vertex for linear approximation of this curve.

Draw a straight line from the starting point S to a certain point on the curve (current point),
It is checked whether the distance between the straight line and point X+ on the way is less than or equal to a predetermined tolerance. If it is below the allowable value.

Advance the current point and make the same determination. When the distance exceeds the allowable amount at a certain current point, the point immediately before that point is set as the apex, and the points from there to the starting point S are expressed by a straight line connecting both points. Then, the same operation is continued using the current point as the starting point.

By this method, the apex Pl of the curve in FIG.

P2 is detected and this curve is 5 as shown in FIG.
-PI-P2-E is expressed by three line segments. but,
Observing FIG. 4, it is clear that this curve is more realistic if it is represented by two line segments than by three tree line segments. Moreover, when a geometric figure is digitally represented, the apex portions are generally rounded as shown in FIG. 4, so the above-mentioned problems are likely to occur.

Furthermore, the conventional method described above requires an extremely large amount of calculation.

〔the purpose〕

An object of the present invention is to provide a vertex detection method that can solve the above problems.

〔composition〕

The apex detection method of the present invention calculates average vectors before and after a point of interest on a curve expressed by a direction code string, calculates an angular difference between the average vectors, and compares the angular difference with a predetermined value. It is characterized in that when the angle difference exceeds a fixed value at the location, the point of interest is detected as the 8th point on the curve.
- An embodiment of the present invention will be described below with reference to one drawing.

Figure 1 is a flowchart of vertex detection.
is the direction code of the nth point of the chain code represented curve. Each direction code is a vector as shown in FIG. 3(b), and expressed as a vector as a linear equation.

1= (1,0) 2= (L L) 3= (0,1) 4= (-1,1) 5=(-L,0) 6= (-ri,-1) 7= (0, -1) 8=(1-,-1)
(1) The processing content of this flowchart is that when the angular difference between the average vectors before and after a point of interest on a curve expressed as a direction code string is large, that point is
rner).

First, initialize the chain element number n to 1.

Processing is started (step 51). Increment n by 1 (step 52), and check whether the direction code Cn is 0 (step 53). If C-n is 0, that is, if the point is not on the curve, the process ends.

If Cn≠O, calculate the average vector Sc from that point (point of interest) to the previous point (step 54), and calculate the vector (direction code) Cn of the point of interest and the previous average vector Sc. Angle difference 1cnAsc1. Specifically, the angle O in the following equation is compared with the set value Th (step 55).
).

If the ICn1·l5nl angular difference is less than the predetermined value, the point of interest is not a vertex, and the process returns to step 52 and continues the process with the primary point as the point of interest.

If the angular difference is larger than the predetermined value, the average vector Ec from the point of interest to the subsequent r points is determined (step 56).

The angular difference 1sc'Ecl between this vector Ec and the average vector Sc before the point of interest is compared with a predetermined value Th (step 57). If the angular difference is less than or equal to the predetermined value, the point of interest is not a vertex, so the process returns to step 52 and proceeds to process the next point of interest.

If the angular difference is larger than the predetermined value, the point of interest is detected as a vertex (step 58), n is incremented by r (step 59), and the process returns to step 53.

The above process cannot be executed by software or hardware, but we will discuss a specific example of that hardware in the second section.
This will be explained using figures.

In this figure, 100 represents seven stages 101 to 10.
It is a shift register consisting of 7 and 6 stages 104 into which a direction code chain is input from the left end of the figure.
The content is the direction code of the point of interest. vector adder 1
10 converts the direction codes (vectors) held in the stages 105, 106, and 107 into the vector expression of equation (1) and adds them to obtain an average vector 9c. The differentiator 112 calculates the average vector Sc and the vector Cn of the point of interest.
The angular difference θ is calculated using equation (2).

The comparator 115 sets the angular difference O to a predetermined value Th.

If θ>Th, the switch 116 is activated to input the output (Sc) of the vector adder 110 and the output of the other vector adder 111 to the difference unit 113. If θ≦Th, the switch 116 is activated, and then one pulse is sent from the pulse generator 113, and the contents of the shift register 100 are shifted to the right by one stage (the point of interest advances by one).

Vector adder 111 includes stages 101, 102.1
The direction code held in 03 is changed to the form of equation (1) and added to obtain the average vector Ec. When the switch 116 is activated and the average vectors Sc and Ec are input to the differentiator 113, an angular difference lsc'Ecl is obtained and compared with a predetermined value Th by the comparator 115. If the angular difference is greater than Th, a vertex detection signal is sent from the comparator 115, and the point of interest is detected as a vertex. When this apex detection signal is generated, the counter 118 operates and the pulse generator 117 outputs three shift pulses.
6, and the contents of the shift register 100 are shifted to the right by three stages 6, that is, the point of interest advances three steps.

By sequentially connecting the vertices detected as described above with straight lines, a curve can be approximated by a polygonal line. for example,
The curve in FIG. 4 has two detected vertices and is approximated by two straight lines.

Next, a method for examining the similarity between two curves by performing approximations of 1 or more will be described.

Using the angle difference θ and the length Q between the vertices, one curve (Q
, θ1. It can be expressed as Q1 θ71”'Qmθm).

Another similarly expressed curve (Q,'θ,'Q?'
0? '! ”・Qm′ θ,′) The degree of similarity between both curves can be defined by calculating the distance with the following formula7...(3) Means the minimum value when changing from p to q do.

〔effect〕

According to the method of the present invention described above, it is possible to detect vertices more accurately than before, and as is clear from the above embodiments, the amount of calculation for detecting vertices is considerably smaller than before. can get.

[Brief explanation of drawings]

FIG. 1 is a flowchart of vertex detection processing in an embodiment of the present invention, FIG. 2 is a block diagram showing an example of hardware that executes the same processing, and FIG. 3 is an explanatory diagram of direction codes.
FIG. 4 is a diagram showing a curve expressed by a chain of direction codes, and FIG. 5 is a diagram showing the result of linear approximation by a conventional method to the curve in FIG. 4. 6 100...shift register, tto, ilt・
...Vector adder, 112.113...Differentiator. 114.115...Comparator, 116...Switch. 117...Pulse generator, 118...Counter. Figure 2 Figure 3 Figure 4 Figure 5

Claims (1)

[Claims] (1) Find the average vectors before and after the point of interest on the curve expressed by the direction code string, find the angular difference between the average vectors, compare the angular difference with a predetermined value, and determine whether the angular difference is A vertex detection method characterized in that when a predetermined value is exceeded, the point of interest is detected as the vertex of the curve.